Create a backup job
A job is one source, one destination and a schedule. The wizard (Jobs → New Job) starts from the task and asks five things — What, From, To, When, Done. The finer options sit in collapsed sections on the step they belong to, and which ones appear depends on the task, so you are never hunting for a switch that is not there for your kind of job.
Each task comes with its own defaults — a nightly file backup at 02:00 with the usual cache and temp exclusions, Microsoft 365 every 6 hours, a daily cloud copy, a weekly archive. Change what you need; nothing is locked in until you save.
Step 1 — What: the task #
The tasks #
Under What do you want to do?, pick what you are trying to do; the wizard then asks only what that task needs.
| Task | What it covers | Runs on |
|---|---|---|
| Back up a computer or server | Folders on a machine with the agent installed. The workhorse. | That machine's agent |
| Back up a whole Windows machine | A bootable image of every disk, to restore as a VM. See Whole machine. | That Windows machine's agent |
| Copy one cloud storage to another | One storage connection copied into another — S3 to Azure, SharePoint to a NAS, whole containers or one prefix. | Your agent (relays the data) or the Caryvane Cloud Agent (data never touches your hardware) |
| Back up Microsoft 365 / Back up Google Workspace | Microsoft 365 or Google Workspace. See Microsoft 365 & Google. | Your agent or the Cloud Agent |
| Back up Hyper-V, SQL Server or Active Directory | Hyper-V VMs, SQL Server databases or Active Directory, captured through their VSS writers so the application is quiesced first. | The Windows machine's agent, only |
| Back up a Proxmox server | Virtual machines on a Proxmox VE cluster, backed up from Proxmox itself. This card opens the cluster's page instead of the next steps. See Proxmox. | The Proxmox cluster |
| Archive old files to cheaper storage | Moves files past their folder's retention rule out of live storage. See Archiving Policy. | The Cloud Agent, only |
| Restore files to a machine | From a backup in storage back onto a machine. See Restore data. | The destination machine's agent |
What a job does and where it runs cannot be changed after the job is created — create a new job, or duplicate the existing one (Jobs → ⋮ → Duplicate).
Customer. When you can see more than one customer, choose whose job this is first; machines and storage are then filtered to that customer. For a job on a machine the customer is implied by the machine; for a Cloud Agent job it must be stated, or the job lands under your own tenant with your own allowance.
Restore. Restoring is a task of its own, Restore files to a machine, for files only; see Restore data.
Step 2 — From #
Files #
Choose the agent, then add source paths — absolute paths on that machine (C:\Data, D:\Shares\Accounts, /srv/files). The folder picker browses the machine live through its agent. Each run sends: Only what changed uploads only files that changed since the last run (by size and modified time, then by content hash when the destination reports one); Everything, every time uploads every file on each run.
Each path is namespaced at the destination by its full source path (C/Data/…, D/Shares/Accounts/…), so two folders that both contain report.docx can never overwrite each other.
Cloud Storage #
Choose the source storage target and, if it runs on your agent, the relay agent (objects are staged on that machine between the two targets). Add one or more container or container/prefix specs — the browse button lists the source. Each run copies Only what changed or Everything, every time, as above.
App-consistent #
For Back up Hyper-V, SQL Server or Active Directory, choose the Windows machine and a workload — Hyper-V virtual machines, SQL Server databases or Active Directory (NTDS, on a domain controller) — then Discover on machine to list what is there and tick the VMs or databases you want. Ticking none means all of them, including ones created later. SQL is captured copy-only, so the customer's own log-backup chain is untouched. A Hyper-V VM on Server 2016 or later can be backed up incrementally (only changed blocks travel) once that is switched on for the agent through the API (how); until then every run is a complete, application-consistent copy of the VM's files.
Step 3 — To #
A storage target of the customer's (Storage), plus an optional directory within it (Folder in that storage). Everything the job writes goes under that directory, so one bucket can hold many jobs. Local folders and SMB shares cannot be the destination of a Cloud Agent job — the cloud fleet has no route to a customer's LAN — and the wizard says so rather than letting the job fail on its first run.
Step 4 — When #
| Frequency | Meaning | Available for |
|---|---|---|
| Scheduled | A 5-field cron expression; presets for daily, weekly, hourly. Times are UTC. | Everything |
| Continuous | Changes are picked up as they happen — the agent watches the filesystem, or the storage provider reports its own changes. | Files jobs; Cloud Agent copies from Azure Blob, S3, Wasabi or Egnyte |
| Manual | Only when someone presses Run. | Everything |
Connector jobs run every N hours instead of on a cron (6 by default).
0 2 * * * runs at 02:00 UTC — 03:00 in British Summer Time. Write the cron for the wall-clock time you want minus the offset, and remember it shifts by an hour twice a year.
The options on each step #
The finer settings, once a step of their own, sit in collapsed sections; collapsed, each shows one line saying what is set. On From: Where it runs (One of your machines or Caryvane Cloud Agent, for the tasks that allow both) and What to include; on To: Existing files and history; on When: Before and after each run. Open a section only when the default is not what you want.
When a file already exists at the destination #
Replace writes the current file over the stored one; Leave it keeps the stored one; Keep both saves the new copy under a new name. Continuous jobs, app-consistent captures, machine images and connector syncs always Replace (each writes the current state over the last), and the wizard shows the choice locked with the reason.
Version history #
Version history is set on the storage (Storage → ⋮ → Version history… → Keep previous versions): the last N copies of every file that is overwritten (3, 5, 10 or a number you choose), optionally for a minimum age as well. A job follows its storage's policy while Use the storage's version history is on; turn that off to give one job a policy of its own. Where the destination has native versioning (S3-compatible buckets with versioning enabled) Caryvane uses it; elsewhere it keeps the copies in a sidecar folder beside the data. Versions are what Version history and the drive's context menu show. Off by default because it costs storage. Versions are only made when the job replaces existing files; with Leave it or Keep both the wizard warns. See Version history.
What to include (Files jobs) #
- Age filter — only files newer than N days, or only files older than N days / not accessed in N days. The older-than modes plus Delete from the machine after upload make a simple archive: cold files are moved out.
- Include open and locked files — reads open and locked files on Windows through a Volume Shadow Copy (VSS). On by default for the nightly template. This is file-level; for a database or VM use Back up Hyper-V, SQL Server or Active Directory instead.
- Delete from the machine after upload — a move, not a copy. Each file is deleted only after it is safely at the destination.
- Leave out — Patterns to leave out as globs (
*.tmp, node_modules/, logs/*.log), with one-click chips for the common ones; the template pre-fills cache and temp junk. Skip hidden files / Skip system files switch on the Windows attributes.
Notifications and scripts #
Mute notifications for this job (on the Done step when you edit a job) overrides the customer's notification settings for one noisy job. Under Before and after each run, the scripts Before the run / After the run (PowerShell, cmd, bash; with a Script timeout) run on the machine around a Files job — stop a service, dump a database, then start it again. Not available for jobs the Cloud Agent runs; there is no machine to run them on.
Step 5 — Done #
Every choice on one page, each line linking back to its step. The job's name is here too, made from your answers and yours to change. Anything that would fail on the server — a missing customer, a source and destination that are the same target, a schedule with no cron — is shown here in red before you can save.
After it is created #
From the job's row: Run now, Run History (every run with files uploaded/skipped/failed, bytes, duration, the run log, and the list of failed files with the reason for each), Browse backup (connector jobs), Edit, Duplicate, Disable, Delete. A run ends Completed, Warning (finished, but some files were skipped — encrypted with EFS, in use, or blocked by antivirus; the log names each) or Failed. Run the first backup by hand and read its log before trusting the schedule.
A backup that has never been restored is a hypothesis. Pick one file and bring it back the day the job is created.
Go deeper #
- Files backup — incremental rules, VSS, exclusions, age filter
- Hyper-V — application-consistent VMs and changed-block incrementals
- SQL Server — copy-only captures and attach-to-restore
- Active Directory — the NTDS writer and DSRM restore
- Cloud-to-cloud copy — agent relay or Cloud Agent, change feeds, continuous
- Continuous backup — the watcher, the debounce, when to use it
- Version history — where versions live and how to get one back